Ok, I was trying to come up with a way to illustrate this over-compression / loudness issue in a way you don't have to BE a sound engineer to understand. I've included some "further reading" links at the end. This is just a development (or pale imitation) of those more comprehensive and far better though-out articles.
What I've done is to grab a few CD's from my collection in a somewhat random fashion. I say somewhat, buy I expected some of them to be good or bad examples, depending on the genre of music and/or their age. It's fairly safe to say that if you're looking for the worst examples of this phenomenon then recent pop releases are the place to be
I've used the first song on each album, mostly out of laziness, but partly because albums are mastered with consistency between songs in mind, so the first song is as good an example as any other on the disc. Where possible, I’ve included the name of the Mastering Engineer.
I've grabbed some statistics for each song, too. I'll try a quick explanation of what these mean..
The stats on each song are in left and right columns, for each side of the stereo mix.
Peak amplitude : The highest signal peak during the song, with the highest possible signal being 0db. Any signal which reads 0db may be clipped, which is a distortion of the waveform that is unpleasant to the ear.
Possibly Clipped : The number of samples in each stereo channel that are clipped. The more, the worse.
Maximum RMS power : RMS is a way of averaging out volume in an attempt to gauge the loudness of a section of music. The maximum RMS gives us an idea of how loud the loudest section of the song is.
Average RMS power : Gives us a very rough idea of the overall loudness throughout the song. This can be thrown by one very quiet section, even if the rest is really loud.
I’m thinking it makes sense to go chronologically, so let’s travel back in time to 1989...
Left Right
Peak Amplitude: -1.49 dB -1.99 dB
Possibly Clipped: 0 0
Maximum RMS Power: -9.07 dB -11.28 dB
Average RMS Power: -16.25 dB -16.52 dB
Have you ever thought of ‘Debaser’ as a quiet song? Well, by today’s standards it is.. It’s fairly safe to say it wouldn’t get released at this volume in today’s loudness-conscious world.
For this song it looks like the Mastering Engineer (ME from here on) was aiming to keep the peaks at around -2db. This leaves a nice, healthy amount of headroom below 0db and is a sensible way to work.
The maximum RMS only reaches around -10db during the loudest sections, but the average is around -17db, which tells us that we have a nicely dynamic master on our hands.
Left Right
Peak Amplitude: -2.06 dB -2.58 dB
Possibly Clipped: 0 0
Maximum RMS Power: -9.01 dB -9.87 dB
Average RMS Power: -22.42 dB -23.59 dB
This is an example of an EXTREMELY quiet CD. This is the one mastered CD I know of that people always notice as being quiet. In fact, I’m not even sure it was mastered, as no mastering credit is listed on the sleeve. An average RMS of -22db is extremely low, even if the song does peak at around -2db. In fact, apart from a few exceptions, most of the peaks are around -5db throughout the song. This is the one CD I looked at that could be said to have been mastered ‘too’ quietly.
Left Right
Peak Amplitude: -0.19 dB -0.2 dB
Possibly Clipped: 0 0
Maximum RMS Power: -8.71 dB -8.8 dB
Average RMS Power: -17.24 dB -16.69 dB
From the album 'Gish'
This is a good example of an early 90's rock CD. It's also a very dynamic master, as anyone familiar with the song would imagine. The RMS averages around 17db, but moves up to around 9db for the loud rock-out sections.
A quick look at the waveform shows us that the peaks that are registering at just slightly below 0db are infrequent events, probably loud snare hits. This master clearly sets out to preserve the occasional peak in the music, while retaining the quiet/loud dynamic which is so important to the song.
Mastered by Chris Blair
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 1 1
Maximum RMS Power: -5.62 dB -5.23 dB
Average RMS Power: -10.02 dB -10.26 dB
This song is an altogether different kind of master. Here, there is only a ~5db difference between the maximum and average RMS, and a peek at the waveform indicates an altogether more aggressively compressed master. However, the song concerned is also less dynamic than some of the previous examples.
We also see our first ‘Possibly clipped’ sample here. Upon closer inspection, the waveform concerned escapes clipping, if only by the skin of it’s teeth.
Overall, this is a loud, but carefully mastered song which takes care to avoid clipping, even if it sails close to the wind at times.
Left Right
Peak Amplitude: -.01 dB 0 dB
Possibly Clipped: 354 109
Maximum RMS Power: -2.89 dB -3.17 dB
Average RMS Power: -9.33 dB -10.03 dB
Now we’re getting into the fun stuff. Air. French band. I’d describe them as… kinda mellow. They don’t rock out a lot. Unless you examine their stats, in which case you’d think they were Sabbath on speed.
A max RMS of -3db. That, my friends, is extremely loud. An average RMS of -10db. The same as the rock number from Radiohead above.
The real fun is in the peaks on this one though. Peaks at 0db. Bad sign. 463 possible clipped samples between the two channels. These ain’t no ‘nearly’ clips either..
That’s 26 seconds into the song. It continues in that vein, too.
This is audio abuse! This song has been deliberately distorted in order to gain an extra db or two of loudness, badly compromising the quality of the audio.
Mastered by John Matousek
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 1 36
Maximum RMS Power: -1.99 dB -1.65 dB
Average RMS Power: -9.72 dB -9.88 dB
More madness here.. Not so much in the clipping department, although there’s a few nasties there, too, if not quite on the same scale as the previous song. This is a good example of insane loudness ; a max RMS of -2db is enough to make your eyes water, and there are sections of the song where this level is maintained.
Left Right
Peak Amplitude: -.13 dB -.12 dB
Possibly Clipped: 0 0
Maximum RMS Power: -3.95 dB -3.28 dB
Average RMS Power: -9.43 dB -9.55 dB
Here’s an example of a loud master that’s been shown a bit of TLC. A maximum peak level of -0.12db shows that care has been taken to avoid clipping, and although the loud sections are heavily compressed, the max RMS of -3.3db shows an element of restraint. Thankfully, the quiet parts are also kept relatively quiet, as shown by the waveform.
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 31 4
Maximum RMS Power: -2.3 dB -3.58 dB
Average RMS Power: -7.28 dB -9.01 dB
Hehe. This one’s a bit crazy now. I’m assuming everyone’s somewhat familiar with Mondo Generator’s sound. It’s heavy. But not this heavy. Insane amounts of compression. Once it really kicks off, it’s -3db RMS all the way. LOUD. No dynamic whatsoever. Oddly, the left side of the mix is significantly louder than the right. Strange..
What I've done is to grab a few CD's from my collection in a somewhat random fashion. I say somewhat, buy I expected some of them to be good or bad examples, depending on the genre of music and/or their age. It's fairly safe to say that if you're looking for the worst examples of this phenomenon then recent pop releases are the place to be
I've used the first song on each album, mostly out of laziness, but partly because albums are mastered with consistency between songs in mind, so the first song is as good an example as any other on the disc. Where possible, I’ve included the name of the Mastering Engineer.
I've grabbed some statistics for each song, too. I'll try a quick explanation of what these mean..
The stats on each song are in left and right columns, for each side of the stereo mix.
Peak amplitude : The highest signal peak during the song, with the highest possible signal being 0db. Any signal which reads 0db may be clipped, which is a distortion of the waveform that is unpleasant to the ear.
Possibly Clipped : The number of samples in each stereo channel that are clipped. The more, the worse.
Maximum RMS power : RMS is a way of averaging out volume in an attempt to gauge the loudness of a section of music. The maximum RMS gives us an idea of how loud the loudest section of the song is.
Average RMS power : Gives us a very rough idea of the overall loudness throughout the song. This can be thrown by one very quiet section, even if the rest is really loud.
I’m thinking it makes sense to go chronologically, so let’s travel back in time to 1989...
Pixies - Debaser 1989
Left Right
Peak Amplitude: -1.49 dB -1.99 dB
Possibly Clipped: 0 0
Maximum RMS Power: -9.07 dB -11.28 dB
Average RMS Power: -16.25 dB -16.52 dB
Have you ever thought of ‘Debaser’ as a quiet song? Well, by today’s standards it is.. It’s fairly safe to say it wouldn’t get released at this volume in today’s loudness-conscious world.
For this song it looks like the Mastering Engineer (ME from here on) was aiming to keep the peaks at around -2db. This leaves a nice, healthy amount of headroom below 0db and is a sensible way to work.
The maximum RMS only reaches around -10db during the loudest sections, but the average is around -17db, which tells us that we have a nicely dynamic master on our hands.
The Breeders - Glorious 1990
Left Right
Peak Amplitude: -2.06 dB -2.58 dB
Possibly Clipped: 0 0
Maximum RMS Power: -9.01 dB -9.87 dB
Average RMS Power: -22.42 dB -23.59 dB
This is an example of an EXTREMELY quiet CD. This is the one mastered CD I know of that people always notice as being quiet. In fact, I’m not even sure it was mastered, as no mastering credit is listed on the sleeve. An average RMS of -22db is extremely low, even if the song does peak at around -2db. In fact, apart from a few exceptions, most of the peaks are around -5db throughout the song. This is the one CD I looked at that could be said to have been mastered ‘too’ quietly.
Smashing Pumpkins - Rhinocerous 1991
Mastered by Howie Weinberg
Mastered by Howie Weinberg
Left Right
Peak Amplitude: -0.19 dB -0.2 dB
Possibly Clipped: 0 0
Maximum RMS Power: -8.71 dB -8.8 dB
Average RMS Power: -17.24 dB -16.69 dB
From the album 'Gish'
This is a good example of an early 90's rock CD. It's also a very dynamic master, as anyone familiar with the song would imagine. The RMS averages around 17db, but moves up to around 9db for the loud rock-out sections.
A quick look at the waveform shows us that the peaks that are registering at just slightly below 0db are infrequent events, probably loud snare hits. This master clearly sets out to preserve the occasional peak in the music, while retaining the quiet/loud dynamic which is so important to the song.
Radiohead - Planet Telex 1994Mastered by Chris Blair
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 1 1
Maximum RMS Power: -5.62 dB -5.23 dB
Average RMS Power: -10.02 dB -10.26 dB
This song is an altogether different kind of master. Here, there is only a ~5db difference between the maximum and average RMS, and a peek at the waveform indicates an altogether more aggressively compressed master. However, the song concerned is also less dynamic than some of the previous examples.
We also see our first ‘Possibly clipped’ sample here. Upon closer inspection, the waveform concerned escapes clipping, if only by the skin of it’s teeth.
Overall, this is a loud, but carefully mastered song which takes care to avoid clipping, even if it sails close to the wind at times.
Air - La Femme D'Argent 1998
Mastered by Nilesh Patel
Mastered by Nilesh Patel
Left Right
Peak Amplitude: -.01 dB 0 dB
Possibly Clipped: 354 109
Maximum RMS Power: -2.89 dB -3.17 dB
Average RMS Power: -9.33 dB -10.03 dB
Now we’re getting into the fun stuff. Air. French band. I’d describe them as… kinda mellow. They don’t rock out a lot. Unless you examine their stats, in which case you’d think they were Sabbath on speed.
A max RMS of -3db. That, my friends, is extremely loud. An average RMS of -10db. The same as the rock number from Radiohead above.
The real fun is in the peaks on this one though. Peaks at 0db. Bad sign. 463 possible clipped samples between the two channels. These ain’t no ‘nearly’ clips either..
That’s 26 seconds into the song. It continues in that vein, too.
This is audio abuse! This song has been deliberately distorted in order to gain an extra db or two of loudness, badly compromising the quality of the audio.
At the Drive-In - Rascuache 1999Mastered by John Matousek
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 1 36
Maximum RMS Power: -1.99 dB -1.65 dB
Average RMS Power: -9.72 dB -9.88 dB
More madness here.. Not so much in the clipping department, although there’s a few nasties there, too, if not quite on the same scale as the previous song. This is a good example of insane loudness ; a max RMS of -2db is enough to make your eyes water, and there are sections of the song where this level is maintained.
Muse - Sunburn 1999
Left Right
Peak Amplitude: -.13 dB -.12 dB
Possibly Clipped: 0 0
Maximum RMS Power: -3.95 dB -3.28 dB
Average RMS Power: -9.43 dB -9.55 dB
Here’s an example of a loud master that’s been shown a bit of TLC. A maximum peak level of -0.12db shows that care has been taken to avoid clipping, and although the loud sections are heavily compressed, the max RMS of -3.3db shows an element of restraint. Thankfully, the quiet parts are also kept relatively quiet, as shown by the waveform.
Mondo Generator - 13th Floor 2000
Mastered by Phil Easter
Mastered by Phil Easter
Left Right
Peak Amplitude: 0 dB 0 dB
Possibly Clipped: 31 4
Maximum RMS Power: -2.3 dB -3.58 dB
Average RMS Power: -7.28 dB -9.01 dB
Hehe. This one’s a bit crazy now. I’m assuming everyone’s somewhat familiar with Mondo Generator’s sound. It’s heavy. But not this heavy. Insane amounts of compression. Once it really kicks off, it’s -3db RMS all the way. LOUD. No dynamic whatsoever. Oddly, the left side of the mix is significantly louder than the right. Strange..
